The wind velocity data with UAV contains more continuous outliers, and the prior distribution of noise statistics is known insufficiently. To improve the precision of wind velocity, an adaptive Kalman filter algorithm with restraining outliers presented in this paper. A compressibility function integrated to new information based on analyzing the Kalman Filter algorithm. According to the of the variance and mean value of new information, the weighting factor adjusted adaptively to ensure the initial properties. Simulation and analysis indicate that the algorithm can reduce the influence of outliers, and ensure the precision. The algorithm can be applied in UAV wind measurement.